1. Field of the Invention
The present invention relates to a visual special effect image or video pair suitable for binocular viewing, and a method and device for generating the visual special effect image or video pair suitable for binocular viewing.
2. Relevant Background
Currently, market-oriented 3D display approach mainly presents two-channel visual content to left and right eyes respectively. The viewer can stably fuse the two-channel visual content and generate three-dimensional sensation. Two eyes have a distance of about 6 cm, and thus see an object from two different positions and views. The difference in position of images projected to two eyes respectively is called binocular disparity. Generally, it is thought that two-channel visual content only have binocular disparity, but no other differences, such as difference in brightness, color and so on.
There are many ways to find correspondence between left and right view images for stereoscopic display. Two regions in correspondence are called corresponding regions for two views. The following are several common ways to find corresponding regions: 1) Depth map based method: Assume a point A in one view with location (mx, my). if the view's depth map and camera parameters are already known, we can find the corresponding point A′ in the other view through formulation
            m      x      ′        =                  m        x            +                        ft          x                Z              ,where f is camera focal length and tx is the distance between two cameras. The location of point A′ in the other view is (mx′, my). Corresponding regions can be found based on corresponding points. 2) Content feature based matching algorithm: firstly, significant features are extracted from the two images. Local features are corner, edge, line, etc. Global features are polygon, image structure descriptor, etc. Then, features are represented by parameters. Finally, match the two images by the feature parameters and find corresponding regions. 3) Greyscale based matching algorithm: images are regarded as 2D signals. Matching relationship is found by calculating the correlation function of the two signals. Thus, corresponding regions are found for two images. The way to find corresponding regions is not limited to the above.
When difference of binocular images is little, our brain can fuse left and right views and have stable texture perception and determinate depth perception. The physiological and psychological study of binocular vision reveals that when the difference of binocular images is large, our brain can't fuse left and right views and thus have unstable texture perception and uncertain depth perception. Especially, when regions in binocular images have opposite brightness contrast polarities, and the absolute value of difference between two brightness contrasts exceeds a critical value, these regions are perceived with fluctuating lustre. This phenomenon is called binocular rivalry. The fluctuating lustre sensation is due to the nonuniform, unstable and time varying brightness sensation of the surface.